Attribute-oriented denazification of fuzzy database tuples with categorical entries

We are investigating the ability to data mine fuzzy tuples, which are often utilized to represent uncertainty about the registered information. We discuss different aspects of fuzzy databases and comment on practical advantages of the model we utilized in our research. Motivated by a well known technique called Attribute-Oriented Induction, which has been developed for summarization of ordinary relational databases, we propose a new heuristic algorithm, allowing attribute-oriented defuzzification of fuzzy database tuples to the form acceptable for many regular (i.e. atomic values based) data mining algorithms. Significant advantages of our approach to defuzzification of fuzzy database tuples include: (1) its intuitive character of fuzzy tuples' interpretation, (2) a unique capability of incorporating background knowledge, implicitly stored in the fuzzy database models in the form of fuzzy similarity relation, directly into the imprecise data interpretation process, (3) transformation of fuzzy tuples to a format easy to process by regular data mining algorithms, and (4) a good scalability for time-efficient treatment of large datasets containing non-atomic, categorical data entries.